LLMCost

Document Summarizer LLM Cost

Batch summarization of long documents. Monthly estimates at 2,000 requests/mo, avg 20,000 input / 1,000 output tokens per request.

#ModelEst. monthly costvs cheapest
1Qwen: Qwen3.7 Flash$1.46—
2DeepSeek: DeepSeek V4.1 Flash (batch)$2.301.6×
3Anthropic: Claude Haiku 5.5 (batch)$2.501.7×
4OpenAI: GPT-6 Luna Pro (batch)$2.501.7×
5OpenAI: GPT-6 Luna (batch)$2.501.7×
6DeepSeek: DeepSeek V4.1 Flash$2.711.9×
7OpenAI: GPT-5 Nano$2.801.9×
8Qwen: Qwen3.8 27B$4.302.9×
9Google: Gemini 2.5 Flash Lite$4.803.3×
10Anthropic: Claude Haiku 5.5$5.003.4×
11OpenAI: GPT-6 Luna Pro$5.003.4×
12OpenAI: GPT-6 Luna$5.003.4×
13Qwen: Qwen3.8 Omni Flash$6.944.8×
14Qwen: Qwen3.8 Flash$6.944.8×
15DeepSeek: DeepSeek V4.1 Flash 1017$7.204.9×
16Meta: Llama 4 Maverick$8.806.0×
17DeepSeek: DeepSeek V4 Flash Vision Exp$9.926.8×
18OpenAI: GPT-5.6 Luna$10.407.1×
19OpenAI: GPT-5.4 Nano$10.507.2×
20DeepSeek: DeepSeek V3.1$11.908.2×
21OpenAI: GPT-5 Mini$14.009.6×
22Google: Gemini 3.5 Flash Lite$17.0011.6×
23Google: Gemini 2.5 Flash$17.0011.6×
24Google: Gemini 3.8 Flash (batch)$18.7512.8×
25Google: Gemini 3.7 Flash (batch)$18.7512.8×
26DeepSeek: DeepSeek V4 Flash 0731$20.2413.9×
27DeepSeek: DeepSeek V4 Pro 0813$30.3620.8×
28Google: Gemini 3.7 Flash$37.5025.7×
29Google: Gemini 3.6 Flash$37.5025.7×
30Amazon: Nova Pro 1.0$38.4026.3×
31OpenAI: GPT-5.4 Mini$39.0026.7×
32Google: Gemini 3.8 Flash$41.2528.3×
33Google: Gemini 4 Argon (batch)$50.0034.2×
34OpenAI: GPT-6.1 Sol Pro (batch)$50.0034.2×
35OpenAI: GPT-6.1 Sol (batch)$50.0034.2×
36Anthropic: Claude Sonnet 5.5 (batch)$50.0034.2×
37OpenAI: GPT-6 Sol Pro (batch)$50.0034.2×
38OpenAI: GPT-6 Sol (batch)$50.0034.2×
39OpenAI: o4 Mini$52.8036.2×
40Mistral: Mistral Large 4$62.7643.0×
41OpenAI: GPT-5$70.0047.9×
42Google: Gemini 2.5 Pro$70.0047.9×
43Google: Nano Banana 2.1$75.0051.4×
44Google: Gemini 3.5 Flash$78.0053.4×
45xAI: Grok 4.7$92.0063.0×
46Qwen: Qwen3.8 Max (0902)$92.0063.0×
47Qwen: Qwen3.8 2.4T A95B$92.0063.0×
48xAI: Grok 4.6$92.0063.0×
49OpenAI: o3$96.0065.8×
50Anthropic: Claude Sonnet 5$100.0068.5×
51Google: Gemini 4 Argon$100.0068.5×
52OpenAI: GPT-6.1 Sol Pro$100.0068.5×
53OpenAI: GPT-6.1 Sol$100.0068.5×
54Anthropic: Claude Sonnet 5.5$100.0068.5×
55OpenAI: GPT-6 Sol Pro$100.0068.5×
56OpenAI: GPT-6 Sol$100.0068.5×
57Anthropic: Claude Opus 5.5 (batch)$100.0068.5×
58OpenAI: GPT-5.6 Terra Pro$104.0071.2×
59Anthropic: Claude Opus 5 (batch)$125.0085.6×
60OpenAI: GPT-5.4$130.0089.0×
61Anthropic: Claude Sonnet 4.5$150.00102.7×
62Qwen: Qwen3.8 Max Prime$184.00126.0×
63OpenAI: GPT-5.6 Sol$200.00137.0×
64Anthropic: Claude Opus 5.5$200.00137.0×
65Anthropic: Claude Opus 5$250.00171.2×
66OpenAI: GPT-6 Astra (batch)$250.00171.2×
67OpenAI: GPT-6 Astra Pro (batch)$250.00171.2×
68Anthropic: Claude Fable 5.1 (batch)$250.00171.2×
69OpenAI: GPT-5.5$260.00178.1×
70Anthropic: Claude Fable 5$500.00342.5×
71OpenAI: GPT-6 Astra$500.00342.5×
72OpenAI: GPT-6 Astra Pro$500.00342.5×
73Anthropic: Claude Fable 5.1$500.00342.5×
74Anthropic: Claude Opus 4.1$750.00513.7×
75OpenAI: GPT-5 Pro$840.00575.3×
76OpenAI: GPT-5.5 Pro$1560.001068.5×

Cost-saving tips for document summarizers

  • Batch APIs (where offered) cut costs ~50% for non-realtime work.
  • Hierarchical summarization (chunk → summarize → merge) beats single huge prompts on cost.
  • Output is short relative to input — prioritize cheap input pricing.

Customize the numbers with the interactive calculator · All scenarios